EXPERIMENTAL AND THERAPEUTIC MEDICINE 20: 427-435, 2020

Bioinformatics analysis reveals meaningful markers and outcome predictors in HBV‑associated hepatocellular carcinoma

LIJIE ZHANG, JOYMAN MAKAMURE, DAN ZHAO, YIMING LIU, XIAOPENG GUO, CHUANSHENG ZHENG and BIN LIANG

Department of Radiology, Hubei Key Laboratory of Molecular Imaging, Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, Hubei 430022, P.R. China

Received April 20, 2019; Accepted December 5, 2019

DOI: 10.3892/etm.2020.8722

Abstract. Hepatocellular carcinoma (HCC) is the most the high degree of connectivity by using Cytoscape software. common type of malignant neoplasm of the liver with high In addition, overall survival (OS) and disease‑free survival morbidity and mortality. Extensive research into the pathology (DFS) analyses were performed using the Expression of HCC has been performed; however, the molecular Profiling Interactive Analysis online database, which revealed mechanisms underlying the development of hepatitis B that the increased expression of all hub were associated virus‑associated HCC have remained elusive. Thus, the present with poorer OS and DFS outcomes. Receiver operating study aimed to identify critical genes and pathways associated characteristic curves were constructed using GraphPad prism with the development and progression of HCC. The expression 7.0 software. The results confirmed that 15 hub genes were profiles of the GSE121248 dataset were downloaded from the able to distinguish HCC form normal tissues. Furthermore, the Omnibus database and the differentially expression levels of three key genes were analyzed in tumor expressed genes (DEGs) were identified. and normal samples of the Human Protein Atlas database. The (GO) and Kyoto Encyclopedia of Gene and Genome present results may provide further insight into the underlying (KEGG) analyses were performed by using the Database mechanisms of HCC and potential therapeutic targets for the for Annotation, Visualization and Integrated Discovery. treatment of this disease. Subsequently, protein‑protein interaction (PPI) networks were constructed for detecting hub genes. In the present study, Introduction 1,153 DEGs (777 upregulated and 376 downregulated genes) were identified and the PPI network yielded 15 hub genes. GO Hepatocellular carcinoma (HCC) is the third leading cause of analysis revealed that the DEGs were primarily enriched in cancer‑associated mortality worldwide and is highly associ- ‘protein binding’, ‘cytoplasm’ and ‘extracellular exosome’. ated with hepatitis C and B infection, as well as alcohol KEGG analysis indicated that DEGs were accumulated in abuse (1). In East Asia and sub‑Saharan Africa, chronic ‘metabolic pathways’, ‘chemical carcinogenesis’ and ‘fatty hepatitis B infection is a vital factor in the pathogenesis of acid degradation’. After constructing the PPI network, liver cancer (2). To date, numerous studies have identified cyclin‑dependent 1, cyclin B1, cyclin A2, mitotic arrest hepatitis B virus (HBV)‑associated factors, including hepatitis deficient 2 like 1, cyclin B2, DNA topoisomerase IIα, budding B e antigen and the hepatitis B surface antigen, as key predic- uninhibited by benzimidazoles (BUB)1, TTK protein kinase, tors of HBV‑associated HCC development in patients with non‑SMC condensin I complex subunit G, NDC80 kinetochore hepatitis B infection (3). In addition, the viral load of HBV, complex component, A, kinesin family member duration of infection and severity of liver disease have been 11, cell division cycle 20, BUB1B and abnormal spindle linked to the risk of developing cirrhosis (2,4); however, the microtubule assembly were identified as hub genes based on mechanisms of HBV‑associated HCC development remain to be fully elucidated, which poses challenges for the clinical treatment of this disease. Post‑operative recurrence is the most imperative issue to overcome (5). Bioinformatics analyses may be applied to determine the potential mechanisms underlying Correspondence to: Professor Bin Liang, Department of Radiology, the development of HBV‑associated HCC. Furthermore, Hubei Key Laboratory of Molecular Imaging, Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, predictive biomarkers may improve the prediction of patient 1277 Jiefang Road, Wuhan, Hubei 430022, P.R. China prognosis and the understanding of the molecular mechanisms E‑mail: [email protected] underlying this disease (6). α‑Fetoprotein, associated with gene duplications (7), was Key words: bioinformatics analysis, hepatocellular carcinoma, the sole biomarker reported in randomized controlled trials (8). biomarker, differentially expressed genes, hepatitis B virus As only a small number of biomarkers have been identified, developments in the treatment and diagnosis of HCC, as well as prognostic evaluation, are restricted. With recent advances 428 ZHANG et al: BIOINFORMATICS ANALYSIS OF HBV-ASSOCIATED HCC in high‑throughput technologies, an increasing number of gene to evaluate and construct a PPI network of target genes and chips are available for the detection of differentially expressed DEGs, including downregulated and upregulated genes, was genes (DEGs); thus, bioinformatics analyses of associated used. Subsequently, Cytoscape software (17) was utilized to data are frequently being performed (9). Based on these select hub genes in the PPI network. In addition, DAVID was novel techniques, databases of vast core gene data have been used to perform GO and KEGG analyses to obtain additional produced, including the National Center for Biotechnology information on the hub genes. Information (NCBI) Gene Expression Omnibus (GEO) and the Cancer Genome Atlas (10). Analysis of these freely avail- Association of hub genes with clinical outcomes and diag‑ able gene expression data may lead to identification of DEGs nostic value. Overall survival (OS) and disease‑free survival in tumor vs. normal tissue, which may provide novel insight (DFS) analyses of hub genes were performed using the Gene into the development of cancers. Jiao et al (11) reported that Expression Profiling Interactive Analysis GEPIA)( web server upregulated eukaryotic translation initiation factor 2B subunit (http://gepia.cancer‑pku.cn/); log‑rank P<0.05 was considered ε (EIF2B5) expression was associated with HCC development to indicate statistical significance. Immunohistochemical based on GSE54236 and GSE76427 microarray data, indi- data for three hub genes with the highest degrees in patients cating that EIF2B5 may be employed as a novel biomarker for with or without HCC were then downloaded from the Human patients with liver cancer. Protein Atlas (HPA; https://www.proteinatlas.org/). In addi- In the present study, the gene expression dataset tion, GraphPad prism 7 software (GraphPad Software, Inc.) GSE121248, including hepatitis B‑induced HCC and adjacent was used to generate receiver operating characteristic (ROC) normal samples, was subjected to bioinformatics analysis. curves to determine the diagnostic value of the 15 hub genes The DEGs were screened and the functions and associated in HCC. pathways in HBV‑associated HCC were further evaluated. The core genes identified in the present study may be considered Results as potential novel targets for the treatment of HBV‑associated HCC. The present results may also improve the understanding Identification of DEGs in HBV‑associated HCC. The dataset of the development and recurrence of this disease, and provide comprised 107 samples, including 70 tumor samples and 37 a basis for developments regarding its clinical treatment. adjacent normal tissue samples. The GEO2R tool was used to identify the DEGs, with adjusted P<0.05 and |logFC|>1 selected Materials and methods as the cut‑off criteria. A total of 1,153 DEGs, comprising 376 downregulated and 777 upregulated genes, were retrieved by Data sourcing and identification of DEGs. The GSE121248 analyzing the GSE121248 dataset. All DEGs were included in microarray dataset was acquired and downloaded from the the subsequent analysis. GEO database (http://www.ncbi.nlm.nih.gov/geo/), on the basis of the GPL570 platform analyzed by the Affymetrix GO functional and KEGG pathway enrichment analyses U133 Plus 2.0 Array, and included the data of DEGs. After the DEGs were identified, DAVID was of 70 HBV‑associated HCC tumor samples and 37 adjacent employed to perform GO functional and KEGG pathway normal tissue samples. Simultaneously, the Series Matrix enrichment analyses of the DEGs. The data of the DEGs File of GSE121248 was downloaded. In order to obtain more were analyzed with the tool; in the GO analysis, the func- meaningful targets for the clinical application, |log fold change tional terms in three different categories, namely cellular (FC)|>1 (12) was set as the limit and GEO2R (https://www. component, biological process and molecular function ncbi.nlm.nih.gov/geo/geo2r/) was used. Genes were identified enriched by the DEGs were determined. In summary, the as significant DEGs based on adjusted P<0.05 (13), which was DEGs were significantly accumulated in cellular component applied to reduce the false‑positive rate. terms including ‘protein binding’ (GO:0005515), ‘cytoplasm’ (GO:0005737), ‘extracellular exosome’ (GO:0070062) and Gene Ontology (GO) functional and Kyoto Encyclopedia of ‘cytosol’ (GO:005829). Specifically, upregulated genes were Genes and Genomes (KEGG) pathway enrichment analysis mainly enriched in cellular component terms, including of DEGs. GO analysis in the categories cellular component, ‘extracellular region’ (GO:0005576), ‘extracellular exosome’ molecular function and biological process (14) was performed (GO:0070062) and ‘extracellular space’ (GO:0005615), while for the functional annotation of genes. KEGG was used downregulated genes were mainly enriched in biological to determine the biological pathways associated with the process terms, including ‘cell division’ (GO:0051301) and DEGs (15). The Database for Annotation, Visualization and ‘mitotic nuclear division’ (GO:0007067). In Table I, the top Integrated Discovery (DAVID; http://david.ncifcrf.gov) (16), a 10 GO terms of the upregulated and downregulated genes free online biological database, was employed to perform GO are listed according to their P‑value from the lowest to the and KEGG analyses with the Bingo plug‑in to obtain system- highest value. atic and comprehensive information for all DEGs. P<0.05 was Furthermore, KEGG pathway analysis revealed that the considered to indicate a statistically significant difference. DEGs were mainly involved in ‘metabolic pathways’ [Homo sapiens (hsa)01100], ‘chemical carcinogenesis’ (hsa05204) Construction of the protein‑protein interaction (PPI) network. and ‘fatty acid degradation’ (hsa00071). In addition, the To investigate the mechanism underlying the development of upregulated genes were mainly enriched in ‘retinol metabo- HCC, the Search Tool for the Retrieval of Interacting Genes lism’ (hsa00830), which is associated with the development (https://string‑db.org/), a free online tool that may be employed of diabetes (18). Downregulated genes were accumulated EXPERIMENTAL AND THERAPEUTIC MEDICINE 20: 427-435, 2020 429

Table I. GO analysis of differentially expressed genes.

A, Upregulated genes

Category Term Count P‑value FDR

GOTERM_CC_DIRECT GO:0005576‑extracellular region 119 3.20x10‑20 4.34x10‑17 GOTERM_CC_DIRECT GO:0070062‑extracellular exosome 166 1.25x10‑18 1.69x10‑15 GOTERM_CC_DIRECT GO:0005615‑extracellular space 99 2.59x10‑16 3.00x10‑13 GOTERM_CC_DIRECT GO:0072562‑blood microparticle 29 1.31x10‑14 1.78x10‑11 GOTERM_BP_DIRECT GO:0055114‑oxidation‑reduction process 58 4.38x10‑14 7.78x10‑11 GOTERM_MF_DIRECT GO:0016705‑ activity, acting on paired donors, with incorporation or reduction of molecular oxygen 18 5.79x10‑13 9.05x10‑10 GOTERM_BP_DIRECT GO:0019373‑epoxygenase P450 pathway 12 6.99x10‑13 1.24x10‑9 GOTERM_MF_DIRECT GO:0020037‑heme binding 25 3.51x10‑12 5.48x10‑9 GOTERM_MF_DIRECT GO:0019825‑oxygen binding 16 4.49x10‑12 7.01x10‑9 GOTERM_CC_DIRECT GO:0031090‑organelle membrane 20 1.06x10‑11 1.44x10‑8

B, Downregulated genes

Category Term Count P‑value FDR

GOTERM_BP_DIRECT GO:0051301‑cell division 38 7.08x10‑20 1.18x10‑16 GOTERM_BP_DIRECT GO:0007067‑mitotic nuclear division 29 6.27x10‑16 1.11x10‑12 GOTERM_CC_DIRECT GO:0000777‑condensed kinetochore 16 4.44x10‑12 5.98x10‑9 GOTERM_BP_DIRECT GO:0007062‑sister chromatid cohesion 17 9.20x10‑12 1.53x10‑8 GOTERM_CC_DIRECT GO:0030496‑midbody 18 1.46x10‑11 1.97x10‑8 GOTERM_BP_DIRECT GO:0000082‑G1/S transition of mitotic cell cycle 16 9.34x10‑11 1.55x10‑7 GOTERM_CC_DIRECT GO:0000775‑chromosome, centromeric region 12 9.48x10‑10 1.28x10‑6 GOTERM_MF_DIRECT GO:0005515‑protein binding 189 5.91x10‑9 8.35x10‑6 GOTERM_BP_DIRECT GO:0000281‑mitotic cytokinesis 9 1.37x10‑8 2.28x10‑5 GOTERM_CC_DIRECT GO:0000776‑kinetochore 12 4.49x10‑8 6.04x10‑5

FDR, false discovery rate; GO, gene ontology; CC, cellular component; BP, biological process; MF, molecular function.

in the pathways of ‘cell cycle’ (hsa04110) and ‘extracellular protein kinase (TTK) (Table III). To further investigate the matrix‑receptor interaction’ (hsa04512), as well as ‘p53 hub genes, a heatmap and box plot of the expression of the signaling pathway’ (hsa04115) (Table II). top 15 hub genes in HBV‑associated HCC tumor samples and adjacent normal tissue samples were generated using PPI network analysis of DEGs. As the minimum interaction GraphPad prism 7 software (Fig. 1). The PPI networks of score required, 0.900 was applied. The PPI network of all upregulated and downregulated genes, as well as hub genes, the DEGs contained 677 nodes and 1,118 edges (enrich- are presented in Fig. 2. ment P<1.0x10 ‑16). Cytoscape software was then employed to identify the top 15 hub genes that may be associated Survival analyses of hub genes. The GEPIA online database, with HBV‑associated HCC according to their high degree a web‑based tool, was designed to rapidly obtain customizable of connectivity from high to low based on the results of functionalities (19). Utilizing this tool, OS (Fig. 3) and the DFS PPI network. These genes included abnormal spindle‑like (Fig. 4) curves were obtained and the log‑rank P‑values of the 15 microcephaly‑associated protein (ASPM), aurora kinase A hub genes were determined. The results indicated that all of the (AURKA), budding uninhibited by benzimidazoles (BUB)1, hub genes except CCNB2 were significantly associated with OS. BUB1B, cyclin A2 (CCNA2), CCNB1, CCNB2, cell divi- Subsequently, the series matrix data downloaded from the GEO sion cycle 20 (CDC20), cyclin‑dependent kinase 1 (CDK1), database were analyzed and ROC curves were generated to gain kinesin family member 11 (KIF11), mitotic arrest deficient a comprehensive understanding of the predictive value of the 2 like 1 (MAD2L1), non‑SMC condensin I complex subunit hub genes. The results demonstrated that all the hub genes were G (NCAPG), NDC80 kinetochore complex component able to efficiently distinguish HCC tissues from normal tissues (NDC80), DNA topoisomerase II α (TOP2A) and TTK (Fig. 5). In addition, the top three genes were selected based on 430 ZHANG et al: BIOINFORMATICS ANALYSIS OF HBV-ASSOCIATED HCC

Table II. KEGG analysis of differentially expressed genes (top Table III. Top 15 hub genes in the protein‑protein interaction 5 according to P‑value). network ranked by degree.

A, Upregulated genes Gene symbol Degree

Term Count P‑value CDK1 90 CCNB1 79 ‑14 hsa01100:Metabolic pathways 107 2.83x10 CCNA2 75 ‑11 hsa05204:Chemical carcinogenesis 21 9.07x10 MAD2L1 73 ‑9 hsa00830:Retinol metabolism 17 7.29x10 CCNB2 70 ‑8 hsa04610:Complement and 17 2.35x10 TOP2A 70 coagulation cascades BUB1 69 ‑8 hsa00980:Metabolism of 17 6.78x10 TTK 67 xenobiotics by cytochrome P450 NCAPG 67 NDC80 66 B, Downregulated genes AURKA 66 KIF11 66 Term Count P‑value CDC20 65 BUB1B 63 ‑15 hsa04110:Cell cycle 21 6.49x10 ASPM 63 hsa04512:ECM‑receptor interaction 11 1.44x10‑6 hsa04115:p53 signaling pathway 9 1.28x10‑5 CDK1, cyclin‑dependent kinase 1; CCNA2, cyclin A2; MAD2L1, hsa05200:Pathways in cancer 20 2.06x10‑5 mitotic arrest deficient 2 like 1; TOP2A, DNA topoisomerase IIα; hsa03030:DNA replication 7 2.36x10‑5 BUB1, budding uninhibited by benzimidazoles 1; TTK, TTK protein kinase; NCAPG, non‑SMC condensin I complex subunit G; NDC80, NDC80 kinetochore complex component; AURKA, aurora kinase A; KEGG, Kyoto Encyclopedia of Genes and Genomes; hsa, Homo KIF11, kinesin family member 11; CDC20, cell division cycle 20; sapiens; ECM, extracellular matrix. ASPM, abnormal spindle microtubule assembly.

the degree of connectivity in the PPI network for analysis of the the kinase inhibitor compounds BA‑12 and B1‑14, which led to immunohistochemical data in the HPA. Additionally, represen- necrosis and apoptosis of tumor cells without toxicity to adja- tative images indicated that the expression of all three hub genes cent normal cells. In addition, the current study reported that were upregulated in HCC tissues (Fig. 6). CDK1 was involved in the p53 signaling pathway. However, the true mechanism underlying the correlation between CDK1 Discussion and HCC remains unclear (22). Previous studies have indicated that the CCNB1‑Cdk1 complex is a key regulator of mitotic Previous studies have investigated biomarkers and therapeutic entry (23,24). Chai et al (24) revealed that CCNB1 is highly targets for HBV‑associated HCC (2); however, the markers expressed in HCC and is closely associated with the poor currently available are unsatisfactory. Thus, it is of importance prognosis of patients with HCC, which was consistent with the to identify additional targets for improving the therapeutic present results. This indicated the plausibility of CCNB1 as a efficacy of treatments for this disease. In the present study, potential therapeutic target for HCC. However, Weng et al (25) the gene expression data of the GSE121248 dataset including demonstrated that there was no significant difference in the gene expression data of 107 samples comprising 70 CCNB1 expression between patients with non‑recurrent HCC HBV‑associated HCC tumor samples and 37 adjacent normal and healthy subjects. The study concluded that the role of tissue samples were analyzed. A total of 15 core genes were CCNB1 overexpression in oncogenesis and the progression of identified, namely CDK1, CCNB1, CCNA2, MAD2L1, HCC remains unclear. Thus, further investigation is required CCNB2, TOP2A, BUB1, TTK, NCAPG, NDC80, AURK1, to determine the association between the expression of CCNB1 KIF11, CDC20, BUB1B and ASPM. The majority of these hub and the development of HCC. Of note, CCNA2 was reported genes were significantly associated with the development and to be associated with the formation of the HBV‑CCNA2 progression of HBV‑associated HCC, among which CDK1, chimeric transcript, which may accelerate the cell cycle and CCNB1 and CCNA2 were the top three hub genes based on result in tumor development (26). In the present study, after their degree in the PPI network. constructing the CCNA2‑associated Kaplan‑Meier curves To date, the association between the expression of CDK1, for OS and DFS, which indicated a significant association CCNB1, CCNA2, KIF11 and CCNB2, and the development of between CCNA2 expression and the prognosis of HCC, it was HCC deserves further investigation. Prevo and Pirovano (20) concluded that high expression levels of CCNA2 are associ- reported that CDK1 is able to mediate the transition from ated with poor prognosis of patients with HCC. Furthermore, G2 phase to mitosis via the CDK1/PI3K/β‑catenin signaling the results of ROC curve analysis demonstrated the ability of pathway. Haider et al (21) assessed the inhibition of CDK1 by CCNA2 to distinguish HCC samples from normal samples. EXPERIMENTAL AND THERAPEUTIC MEDICINE 20: 427-435, 2020 431

Figure 1. (A) Volcano map of 15 hub genes in the protein‑protein interaction network according to the highest degree. (B) Expression of various genes in the tumor samples and adjacent normal samples, including ASPM (P<0.0001), AURKA (P<0.0001), BUB1 (P<0.0001), BUB1B (P<0.0001), CCNA2 (P<0.0001), CCNB1 (P<0.0001), CCNB2 (P=0.1067), CDC20 (P<0.0001), CDK1 (P<0.0001), KIF11 (P<0.0001), MAD2L1 (P<0.0001), NCAPG (P<0.0001), NDC80 (P<0.0001), TOP2A (P<0.0001) and TTK (P<0.0001). CDK1, cyclin‑dependent kinase 1; CCNA2, cyclin A2; MAD2L1, mitotic arrest deficient 2 like 1; TOP2A, DNA topoisomerase IIα; BUB1, budding uninhibited by benzimidazoles 1; TTK, TTK protein kinase; NCAPG, non‑SMC condensin I complex subunit G; NDC80, NDC80 kinetochore complex component; AURKA, aurora kinase A; KIF11, kinesin family member 11; CDC2, cell division cycle 20; ASPM, abnormal spindle microtubule assembly; Adj, adjacent.

Figure 2. (A) PPI network of upregulated genes, including 386 nodes and 1,105 edges with a PPI enrichment P<1.0x10‑16. (B) PPI network of 15 hub genes. (C) PI network of downregulated genes, including 297 nodes and 3,910 edges with a PPI enrichment P<1.0x10‑16. PPI, protein‑protein interaction.

KIF11 serves an essential role in centrosome and chromosome as well as ‘spindle formation’ and ‘carcinogenesis’; however, dynamics in mitosis (27). To the best of our knowledge, the KEGG pathway enrichment analysis did not suggest any asso- association between the expression levels of KIF11 and the ciation between a particular pathway and KIF11. Furthermore, development of HBV‑associated HCC has remained elusive. the Kaplan‑Meier curves for OS and DFS suggested that In the GO analysis of the present study, KIF11 was mainly upregulated KIF11 expression was linked to poor prognosis enriched in the terms ‘cytoplasm’, ‘cytosol’ and ‘cell division’, in HBV‑associated HCC. Chen et al (28) identified that the 432 ZHANG et al: BIOINFORMATICS ANALYSIS OF HBV-ASSOCIATED HCC

Figure 3. Overall survival curves for 15 hub genes. (A) CDK1, (B) CCNB1, (C) CCNA2, (D) MAD2L1, (E) CCNB2, (F) TOP2A, (G) BUB1, (H) TTK, (I) NCAPG, (J) NDC80, (K) AURKA, (L) KIF11, (M) CDC20, (N) BUB1B and (O) ASPM. CDK1, cyclin‑dependent kinase 1; CCNA2, cyclin A2; MAD2L1, mitotic arrest deficient 2 like 1; TOP2A, DNA topoisomerase IIα; BUB1, budding uninhibited by benzimidazoles 1; TTK, TTK protein kinase; NCAPG, non‑SMC condensin I complex subunit G; NDC80, NDC80 kinetochore complex component; AURKA, aurora kinase A; KIF11, kinesin family member 11; CDC20, cell division cycle 20; ASPM, abnormal spindle microtubule assembly; TPM, transcripts per million.

Figure 4. Disease‑free survival curves of 15 hub genes. (A), CDK1, (B) CCNB1, (C) CCNA2, (D) MAD2L1, (E) CCNB2, (F) TOP2A, (G) BUB1, (H) TTK, (I) NCAPG, (J) NDC80, (K) AURKA, (L) KIF11, (M) CDC20, (N) BUB1B and (O) ASPM. CDK1, cyclin‑dependent kinase 1; CCNA2, cyclin A2; MAD2L1, mitotic arrest deficient 2 like 1; TOP2A, DNA topoisomerase IIα; BUB1, budding uninhibited by benzimidazoles 1; TTK, TTK protein kinase; NCAPG, non‑SMC condensin I complex subunit G; NDC80, NDC80 kinetochore complex component; AURKA, aurora kinase A; KIF11, kinesin family member 11; CDC20, cell division cycle 20; ASPM, abnormal spindle microtubule assembly; TPM, transcripts per million. EXPERIMENTAL AND THERAPEUTIC MEDICINE 20: 427-435, 2020 433

Figure 5. Receiver operating characteristics curves for the 15 hub genes to identify HCC tissue from normal tissue. (A) CDK1 (AUC=0.9282, P<0.0001), (B) CCNB1 (AUC=0.0.9019, P<0.0001), (C) CCNA2 (AUC=0.8973, P<0.0001), (D) MAD2L1 (AUC=0.8514, P<0.0001), (E) CCNB2 (AUC=0.5849, P=0.1496), (F) TOP2A (AUC=0.9525, P<0.0001), (G) BUB1 (AUC=0.9174, P<0.0001), (H) TTK (AUC=0.9143, P<0.0001), (I) NCAPG (AUC=0.922, P<0.0001), (J) NDC80 (AUC=0.9363, P<0.0001), (K) AURKA (AUC=0.8965, P<0.0001), (L) KIF11 (AUC=0.8819, P<0.0001), (M) CDC20 (AUC=0.9189, P<0.0001), (N) BUB1B (AUC=0.9332, P<0.0001) and (O) ASPM (AUC=0.9432, P<0.0001). CDK1, cyclin‑dependent kinase 1; CCNA2, cyclin A2; MAD2L1, mitotic arrest deficient 2 like 1; TOP2A, DNA topoisomerase IIα; BUB1, budding uninhibited by benzimidazoles 1; TTK, TTK protein kinase; NCAPG, non‑SMC condensin I complex subunit G; NDC80, NDC80 kinetochore complex component; AURKA, aurora kinase A; KIF11, kinesin family member 11; CDC20, cell division cycle 20; ASPM, abnormal spindle microtubule assembly; AUC, area under the curve.

Figure 6. Representative histological images from the Human Protein Atlas database (https://www.proteinatlas.org/). The normal liver tissue with staining for CDK1 was from a female subject aged 29 years (patient ID: 1899; staining: Not detected; intensity: Negative; quantity: Negative; location: None; magnification, not available) and the HCC tissue was from a female patient aged 52 years (patient ID: 2399; staining: Medium; intensity: Strong; quantity: <25%; location: Cytoplasmic/membranous nuclear; magnification, not available). The normal liver tissue with staining for CCNB1 was from a female subject aged 32 years (patient ID: 1846; staining: Not detected; intensity: Negative; quantity: Negative; location: None; magnification, not available) and the HCC tissue was from a male patient aged 49 years (patient ID: 929; staining: Medium; intensity: Strong; quantity: <25%; location: Cytoplasmic/membranous; magnification, not available). The normal liver tissue with staining for CCNA2 was from a female subject aged 50 years (patient ID: 2251; staining: Not detected; intensity: Negative; quantity: Negative; location: None; magnification, not available) and the HCC tissue was from a female patient aged 73 years (patient ID: 5192; staining: Low; intensity: Moderate; quantity: <25%; location: Nuclear; magnification, not available). CDK1, cyclin‑dependent kinase 1; CCNA2, cyclin A2; HCC, hepatocellular carcinoma.

overexpression of KIF11 was significantly associated with neoplasms. Conversely, to the best of our knowledge, the shorter relapse‑free survival times; however, the function of biological function of CCNB2 in HCC is largely unknown. KIF11 and the mechanism involved in HCC remains unknown. In the present study, low expression levels of CCNB2 were In addition, Li et al (29) reported that downregulated CCNB2 associated with improved clinical outcomes of patients with expression led to inhibition of the progression of malignant HBV‑associated HCC, including DFS. 434 ZHANG et al: BIOINFORMATICS ANALYSIS OF HBV-ASSOCIATED HCC

The roles of MAD2L1, TOP2A, BUB1, TTK, NCAPG, association between HBV infection and the extent of HCC NDC80, AURKA, CDC20, BUB1B and ASPM in HCC have pathological severity. been well reported in previous studies. As for MAD2L1, In conclusion, the present bioinformatics analysis of Li et al (30) identified this gene as a vital mediator of pathways a microarray dataset identified 15 core genes involved in underlying chromosomal regulation in HCC. Yun et al (31) the development and progression of HBV‑associated HCC. demonstrated that inhibition of MAD2L1 expression led to the Assessment of the association of these key genes with suppression of HCC cell proliferation, migration and invasion, clinical outcomes indicated that these genes may be potential suggesting that MAD2L1 may serve as a potential target for the therapeutic targets for HCC, which may contribute to the clinical treatment and prognostic evaluation of patients with development of treatments for HCC. HCC. In agreement with a previous study (32), TOP2A may be a valuable prognostic marker and predictor of poor survival in Acknowledgements patients with HCC. Based on previous studies and the present results, it was concluded that TOP2A may be a potential Not applicable. biomarker linked to the development of HBV‑associated HCC. BUB1 is a serine/threonine kinase that binds centromeres Funding during mitosis (33), and has been associated with the cell cycle and apoptosis (34), as well as reduced OS in patients The present study was supported by the National Natural with HCC (35). The expression levels of protein kinase human Science Foundation of China (grant nos. 81471765 and monopolar, also known as TTK, were previously reported to 81771950). be markedly increased in HBV‑associated HCC (36); TTK was reported to predict poorer outcomes for patients with Availability of data and materials HBV‑associated HCC (37). The first TTK inhibitors to be devel- oped in combination with taxane chemotherapy have entered The datasets used and/or analyzed during the present study are phase 1 dose‑escalation studies (38). NCAPG, which mediates available from the corresponding author on request. the coiling topology of individual chromatids (39), was proposed to be associated with cell growth, proliferation and migration Authors' contributions in HCC (40). NDC80 was reported to be associated with the progression of HCC (41). Of note, downregulation of NDC80 CZ and BL designed the current study. YL, XG, LZ and DZ was able to suppress the replication of HBV‑associated HCC acquired and analyzed the data. LZ and JM wrote and revised cells (42). Bound to MYC, AURKA, which has two functional the manuscript and analyzed the data. All authors read and non‑synonymous polymorphisms (lle31Phe and Val57lle), was approved the final manuscript. able to regulate tumor cell growth at the genetic level (43). It has been reported that AURKA lle31Phe enhances the proneness Ethics approval and consent to participate of HBV‑infected individuals to develop liver cancer (44). The expression levels of CDC20 have been indicated to be linked Not applicable. to the development of liver cancer and the prognosis of affected patients (45). Thus, it may be a candidate prognostic factor in Patient consent for publication patients with HBV‑associated HCC (46). BUB1B was reported to be involved in the regulation of the cell cycle of tumor Not applicable. cells (47). As for ASPM, its upregulation has been suggested to serve as a novel marker for predicting the progression of HCC, Competing interests early tumor recurrence and poor prognosis (48). Of note, the present study had certain limitations. Data The authors declare that they have no competing interests. sourcing was performed using only one dataset, which may not be sufficient to provide a convincing hypothesis. Analyses References with multiple datasets should be considered in the future. Hepatitis B infection is considered an independent factor in 1. Ghouri YA, Mian I and Rowe JH: Review of hepatocellular carcinoma: Epidemiology, etiology, and carcinogenesis. J the development of HCC in the Cancer Genome Atlas data- Carcinog 16: 1, 2017. base (TCGA) and therefore, TCGA‑based research should be 2. Forner A, Reig M and Bruix J: Hepatocellular carcinoma. conducted in the future. Furthermore, the sample size was Lancet 391: 1301‑1314, 2018. relatively small, which may result in unavoidable bias. Finally, 3. Geier A, Gartung C and Dietrich CG: Hepatitis B e antigen and the risk of hepatocellular carcinoma. N Engl J Med 347: 1721‑1722, 2002. the gene chip used in the present study did not contain any 4. European Association for the Study of the Liver. Electronic address: information on the pathological sections of HBV‑associated [email protected]; European Association for the Study of HCC, and therefore, it was not possible to determine the asso- the Liver: EASL Clinical Practice Guidelines: Management of hepatocellular carcinoma. J Hepatol 69: 182‑236, 2018. ciation between HBV infection and HCC pathological grading 5. Fu PY, Hu B, Ma XL, Yang ZF, Yu MC, Sun HX, Huang A, sample by sample. It is indicated that HBV infection, even Zhang X, Wang J, Hu ZQ, et al: New insight into BIRC3: A novel if temporary, may have a role in promoting the development prognostic indicator and a potential therapeutic target for liver cancer. J Cell Biochem 120: 6035‑6045, 2019. of liver cancer; at the same time, a longer duration of HBV 6. Long J, Zhang L, Wan X, Lin J, Bai Y, Xu W, Xiong J and infection may accelerate the process of HCC (49). However, Zhao H: A four‑gene‑based prognostic model predicts overall survival in patients with hepatocellular carcinoma. J Cell Mol to the best of our knowledge, no biomarker can predict the Med 22: 5928‑5938, 2018. EXPERIMENTAL AND THERAPEUTIC MEDICINE 20: 427-435, 2020 435

7. Sauzay C, Petit A, Bourgeois AM, Barbare JC, Chauffert B, 30. Li Y, Bai W and Zhang J: MiR‑200c‑5p suppresses proliferation Galmiche A and Houessinon A: Alpha‑foetoprotein (AFP): A and metastasis of human hepatocellular carcinoma (HCC) via multi‑purpose marker in hepatocellular carcinoma. Clin Chim suppressing MAD2L1. Biomed Pharmacother 92: 1038‑1044, 2017. Acta 463: 39‑44, 2016. 31. Yun MY, Kim SB, Park S, Han CJ, Han YH, Yoon SH, Kim SH, 8. Tsuchiya N, Sawada Y, Endo I, Saito K, Uemura Y and Kim CM, Choi DW, Cho MH, et al: Mutation analysis of Nakatsura T: Biomarkers for the early diagnosis of hepatocel- p31comet gene, a negative regulator of Mad2, in human hepato- lular carcinoma. World J Gastroenterol 21: 10573‑10583, 2015. cellular carcinoma. ExperimeExp Mol Med 39: 508‑513, 2007. 9. Yan P, He Y, Xie K, Kong S and Zhao W: In silico analyses 32. Wong N, Yeo W, Wong WL, Wong NL, Chan KY, Mo FK, Koh J, for potential key genes associated with gastric cancer. PeerJ 6: Chan SL, Chan AT, Lai PB, et al: TOP2A overexpression in hepa- e6092, 2018. tocellular carcinoma correlates with early age onset, shorter patients 10. Teufel A: Bioinformatics and database resources in hepatology. J survival and chemoresistance. Int J Cancer 124: 644‑652, 2009. Hepatol 62: 712‑719, 2015. 33. Xu B, Xu T, Liu H, Min Q, Wang S and Song Q: MiR‑490‑5p 11. Jiao Y, Fu Z, Li Y, Meng L and Liu Y: High EIF2B5 mRNA suppresses cell proliferation and invasion by targeting BUB1 in expression and its prognostic significance in liver cancer: A hepatocellular carcinoma cells. Pharmacology 100: 269‑282, 2017. study based on the TCGA and GEO database. Cancer Manag 34. Ricke RM, Jeganathan KB and van Deursen JM: Bub1 overex- Res 10: 6003‑6014, 2018. pression induces aneuploidy and tumor formation through Aurora 12. Shen S, Kong J, Qiu Y, Yang X, Wang W and Yan L: B kinase hyperactivation. J Cell Biol 193: 1049‑1064, 2011. Identification of core genes and outcomes in hepatocellular 35. Chen QF, Xia JG, Li W, Shen LJ, Huang T and Wu P: Examining carcinoma by bioinformatics analysis. J Cell Biochem 120: the key genes and pathways in hepatocellular carcinoma develop- 10069‑10081, 2019. ment from hepatitis B virus‑positive cirrhosis. Mol Med Rep 18: 13. Zhou SL, Hu ZQ, Zhou ZJ, Dai Z, Wang Z, Cao Y, Fan J, 4940‑4950, 2018. Huang XW and Zhou J: miR‑28‑5p‑IL‑34‑macrophage feed- 36. Liu X, Liao W, Yuan Q, Ou Y and Huang J: TTK activates Akt back loop modulates hepatocellular carcinoma metastasis. and promotes proliferation and migration of hepatocellular carci- Hepatology 63: 1560‑1575, 2016. noma cells. Oncotarget 6: 34309‑34320, 2015. 14. Gaudet P, Skunca N, Hu JC and Dessimoz C: Primer on the gene 37. Baffy G: Decoding multifocal hepatocellular carcinoma: An ontology. Methods Mol Biol 1446: 25‑37, 2017. opportune pursuit. Hepatobiliary Surg Nutr 4: 206‑210, 2015. 15. Wixon J and Kell D: The Kyoto encyclopedia of genes and 38. Zaman GJR, de Roos JADM, Libouban MAA, Prinsen MBW, genomes‑KEGG. Yeast 17: 48‑55, 2000. de Man J, Buijsman RC and Uitdehaag JCM: TTK inhibitors as 16. Huang da W, Sherman BT and Lempicki RA: Systematic and a targeted therapy for CTNNB1 (β‑catenin) mutant cancers. Mol integrative analysis of large gene lists using DAVID bioinfor- Cancer Ther 16: 2609‑2617, 2017. matics resources. Nat Protoc 4: 44‑57, 2009. 39. Liu W, Liang B, Liu H, Huang Y, Yin X, Zhou F, Yu X, Feng Q, 17. Baryshnikova A: Exploratory analysis of biological networks Li E, Zou Z and Wu L: Overexpression of non‑SMC condensin through visualization, clustering, and functional annotation in I complex subunit G serves as a promising prognostic marker cytoscape. Cold Spring Harb Protoc 2016, 2016. and therapeutic target for hepatocellular carcinoma. Int J Mol 18. McClinton KJ, Aliani M, Kuny S, Sauvé Y and Suh M: Med 40: 731‑738, 2017. Differential effect of a carotenoid‑rich diet on retina function in 40. Liu K, Li Y, Yu B, Wang F, Mi T and Zhao Y: Silencing non‑SMC non‑diabetic and diabetic rats. Nutr Neurosci 11: 1‑11, 2019. chromosome‑associated polypeptide G inhibits proliferation 19. Tang Z, Li C, Kang B, Gao G, Li C and Zhang Z: GEPIA: A and induces apoptosis in hepatocellular carcinoma cells. Can J web server for cancer and normal gene expression profiling and Physiol Pharmacol 96: 1246‑1254, 2018. interactive analyses. Nucleic Acids Res 45: W98‑W102, 2017. 41. Ju LL, Chen L, Li JH, Wang YF, Lu RJ, Bian ZL and Shao JG: 20. Prevo R, Pirovano G, Puliyadi R, Herbert KJ, Effect of NDC80 in human hepatocellular carcinoma. World J Rodriguez‑Berriguete G, O'Docherty A, Greaves W, Gastroenterol 23: 3675‑3683, 2017. McKenna WG and Higgins GS: CDK1 inhibition sensitizes 42. Liu B, Yao Z, Hu K, Huang H, Xu S, Wang Q, Yang Y and Ren J: normal cells to DNA damage in a cell cycle dependent manner. ShRNA‑mediated silencing of the Ndc80 gene suppress cell Cell Cycle 17: 1513‑1523, 2018. proliferation and affected hepatitis B virus‑related hepatocellular 21. Haider C, Grubinger M, Řezníčková E, Weiss TS, Rotheneder H, carcinoma. Clin Res Hepatol Gastroenterol 40: 297‑303, 2016. Miklos W, Berger W, Jorda R, Zatloukal M, Gucky T, et al: 43. Zhang L, Huang Y, Ling J, Zhuo W, Yu Z, Shao M, Luo Y and Novel inhibitors of cyclin‑dependent combat hepatocel- Zhu Y: Screening and function analysis of hub genes and path- lular carcinoma without inducing chemoresistance. Mol Cancer ways in hepatocellular carcinoma via bioinformatics approaches. Ther 12: 1947‑1957, 2013. Cancer Biomark 22: 511‑521, 2018. 22. Zhou Z, Li Y, Hao H, Wang Y, Zhou Z, Wang Z and Chu X: 44. Bao Z, Lu L, Liu X, Guo B, Zhai Y, Li Y, Wang Y, Xie B, Ren Q, Screening Hub genes as prognostic biomarkers of hepatocel- Cao P, et al: Association between the functional polymorphism lular carcinoma by bioinformatics analysis. Cell Transplant 11: Ile31Phe in the AURKA gene and susceptibility of hepatocellular 963689719893950, 2019. carcinoma in chronic hepatitis B virus carriers. Oncotarget 8: 23. Nakayama Y and Yamaguchi N: Role of cyclin B1 levels in DNA 54904‑54912, 2017. damage and DNA damage‑induced senescence. Int Rev Cell Mol 45. Li J, Gao JZ, Du JL, Huang ZX and Wei LX: Increased CDC20 Biol 305: 303‑337, 2013. expression is associated with development and progression of 24. Chai N, Xie HH, Yin JP, Sa KD, Guo Y, Wang M, Liu J, Zhang XF, hepatocellular carcinoma. Int J Oncol 45: 1547‑1555, 2014. Zhang X, Yin H, et al: FOXM1 promotes proliferation in human 46. Chen PF, Li QH, Zeng LR, Yang XY, Peng PL, He JH and Fan B: hepatocellular carcinoma cells by transcriptional activation of A 4‑gene prognostic signature predicting survival in hepatocel- CCNB1. Biochem Biophys Res Commun 500: 924‑929, 2018. lular carcinoma. J Cell Biochem 120: 9117‑9124, 2019. 25. Weng L, Du J, Zhou Q, Cheng B, Li J, Zhang D and Ling C: 47. Wen DY, Lin P, Pang YY, Chen G, He Y, Dang YW and Yang H: Identification of cyclin B1 and Sec62 as biomarkers for recur- Expression of the long intergenic non‑protein coding RNA 665 rence in patients with HBV‑related hepatocellular carcinoma (LINC00665) gene and the cell cycle in hepatocellular carcinoma after surgical resection. Mol Cancer 11: 39, 2012. using the cancer genome atlas, the gene expression omnibus, 26. Chiu YT, Wong JK, Choi SW, Sze KM, Ho DW, Chan LK, Lee JM, and quantitative real‑time polymerase chain reaction. Med Sci Man K, Cherny S, Yang W, et al: Novel pre‑mRNA splicing of Monit 24: 2786‑2808, 2018. intronically integrated HBV generates oncogenic chimera in hepa- 48. Lin SY, Pan HW, Liu SH, Jeng YM, Hu FC, Peng SY, Lai PL and tocellular carcinoma. J Hepatol 64: 1256‑1264, 2016. Hsu HC: ASPM is a novel marker for vascular invasion, early 27. Jin Q, Dai Y, Wang Y, Zhang S and Liu G: High kinesin family recurrence, and poor prognosis of hepatocellular carcinoma. Clin member 11 expression predicts poor prognosis in patients with Cancer Res 14: 4814‑4820, 2008. clear cell renal cell carcinoma. J Clin Pathol 72: 354‑362, 2019. 49. Okuda K, Nakashima T, Sakamoto K, Ikari T, Hidaka H, Kubo Y, 28. Chen J, Li S, Zhou S, Cao S, Lou Y, Shen H, Yin J and Li G: Sakuma K, Motoike Y, Okuda H and Obata H: Hepatocellular Kinesin superfamily protein expression and its association with carcinoma arising in noncirrhotic and highly cirrhotic livers: A progression and prognosis in hepatocellular carcinoma. J Cancer comparative study of histopathology and frequency of hepatitis Res Ther 13: 651‑659, 2017. B markers. Cancer 49: 450‑455, 1982. 29. Li R, Jiang X, Zhang Y, Wang S, Chen X, Yu X, Ma J and Huang X: Cyclin B2 overexpression in human hepatocellular This work is licensed under a Creative Commons carcinoma is associated with poorprognosis. Arch Med Res 50: Attribution-NonCommercial-NoDerivatives 4.0 10‑17, 2019. International (CC BY-NC-ND 4.0) License.